the association experiment - eva.mpg.de filebare peripheral (bper) zero-marked obliques and other...

The Association Experiment

Badut minum buku

prefixing suffixing bothbothisolating

Morphological Typology (from WALS)

isolating

Morphological Typology (from WALS)

Semantic Compositionality

When signs are put together …What is the resulting meaning?

'to bicycle shop'

'bicycle lane'

Semantic Compositionality

A ( X )'entity associated with X'in most languages,observable in genitive construction

Monadic Association Operator

The Association Operator

'entity associated with X and Y'A ( X, Y )

in most languages,a default rule for compositional semantics

Polyadic Association Operator


'entity associated with bicycle and thataway'

BICYCLE THATAWAY

A ( BICYCLE, THATAWAY )


• phylogeneticallyin the language of captive great apes

Gil, David (2005) "Early Human Language was Isolating-Monocategorial-Associational",Proceedings of EVOLANG 6, World Scientific, Singapore, 91-98.

Where the Association Operator is Found

• phylogenetically

Bonobo (Kanzi)using lexigrams

Greenfield and Savage-Rumbaugh (1990)

Orangutan (Chantek)using ASL

Miles (1990)

LIZ HIDE

HIDE AUSTIN

WATER HIDE

HIDE PEANUT

YOU PULL

COME CHANTEK

BEARD PULL

PULL BEARD

in the language of captive great apes


• phylogenetically• ontogenetically

in the language of young infants

Gil, David (2008) "The Acquisition of Syntactic Categories in Jakarta Indonesian",Studies in Language.


• phylogenetically• ontogenetically

in the language of young infants


Hurt kneeHurt truck

Allison 1;8

(Bloom 1973)

pig is hurt by sharp corner of truck][playing with toy pig inside toy truck;

• phylogenetically• ontogenetically• grammatically

as a universal design feature of language


tilfanti leavrahamtilfanti leavrahamxxx-telephone xxx-AbrahamA ( TELEPHONE, ABRAHAM )'entity associated with telepone and Avraham'* 'Beavers build dams''I telephoned Abraham'


• phylogenetically• ontogenetically• grammatically• typologically

languages vary with respect to the extent to whichassociational semantics is supplemented withadditional rules of compositional semantics

fewersuch rules

moresuch rules

highlyassociational

languages

highlyarticulatedlanguages

Ayam makanCHICKEN EAT

'The chicken is eating'

'Someone is eating the chicken''Someone is eating with the chicken''Someone is eating because of the chicken'...'The chicken that is eating''Where the chicken is eating''Why the chicken is eating'...

Riau Indonesian

Ayam makanCHICKEN EAT

'entity associated with chicken and eat'

CHICKEN EAT

A ( CHICKEN, EAT )

Riau Indonesian

But to what extentare other languages like

Riau Indonesian?

Goal: measuring the availability ofapparently associational interpretations

Languages studied: isolatingapparent SVO word order

Semantic domain studied: thematic roles


interpretations that are not obtained bythe application of construction-specific rules,and which therefore may plausibly be characterizedas resulting from the application of theassociation operator

Constructions sought:

Bare Patient Preceding (BPatP)apparent OV word order


Apparently associational interpretationsinvolving thematic roles:

Bare Peripheral (BPer)zero-marked obliques andother more extraneous thematic roles

Testing for BPer

test picture alternative picture

The clown is drinking the book

stimulus 15

4 possible responses

(following, extraneous)

Testing for BPer

Badut minum buku

stimulus 15


?


Testing for BPer

רפסה האת שות הליצן

stimulus 15


?


Testing for BPer

Anh hề uống sách

stimulus 15


?


Testing for BPer

小丑飲書

stimulus 15


?


Testing for BPer

Kitenga tchì ǂxanù

stimulus 15


?


Testing for BPer (preceding, extraneous)

The money is happy

stimulus 8

alternative picture test picture

?

Testing for BPer (preceding, extraneous)

Duit gembira

stimulus 8


?

Testing for BPer

The soldier is cutting the axe

stimulus 14


?

(following, oblique)

Testing for BPer

Tentara potong kampak

stimulus 14


?

(following, oblique)

Testing for BPer (preceding, oblique)

The chairs are jumping

stimulus 6


?

Testing for BPer (preceding, oblique)

Kursi loncat

stimulus 6


?

Testing for BPatP

The dog is drawing

stimulus 3

text picture alternative picture

?

Testing for BPatP

Anjing lukis

stimulus 3


?

Testing for BPatP (plus Ag)

The car is pushing the woman

stimulus 11


?

Testing for BPatP (plus Ag)

Mobil dorong cewek

stimulus 11


?

TOTAL 1575 1488

English 63 63Hebrew 66 66

Papiamentu 46 45Sranan 63 60Bislama 50 47

Twi 60 55Fongbe 48 45Yoruba 66 62

Vietnamese 88 80

Standard Indonesian 133 123Kuala Lumpur Malay 76Kuching Malay 32 31

78

Sundanese 44 39

Riau Indonesian 169 164

Jakarta Indonesian 99104

Minangkabau 77 71

Siak Malay 83 75

Bengkulu 35 32Papua Malay 71 65

Meyah 67 63

Ju|'hoan 3033

Cantonese 61 60

Mentawai 38 37

Numbers of Subjects (as of 4.4.2008)

tota

l

pass

eddi

str/s

tota

l

pass

eddi

str/s

0102030405060708090

100

EnglishHebrewPapiam...SrananBislamaTwiFongbeYorubaJu|'hoanCanton...Vietn...MeyahMentawaiSundaMinangJakarta ISiak MRiau IPapua MBengkuluKucing MKL MStand I

Experimental ResultsBPerBPatP

Conclusions I

• languages range fromhighly articulated to highly associational

• highly associational languagespresent us with a clear view of an evolutionary fossil

Question

• why do languages vary fromhighly articulated to highly associational?

• within the typology of isolating languagesRiau Indonesian is mid-range

Calculated on a representative text by counting:

and then calculating the ratio:

morphemes / concepts

the number of morphemesthe number of concepts belonging tomajor ontological types (thing, property, activity)

••

The Articulation Index

A numerical measure of how muchobligatory grammatical marking a language has

Riau Indonesian

Bislama

Papiamentu

English

Anjing lukis

Dog i dro

E kachó ta pinta

The dog is drawing

2

3

4

6

morphemes

2

2

2

2

concepts

1.0

1.5

2.0

3.0

articu

lation index

Sample calculation of articulation index(based on stimulus 3)


Riau Indonesian Anjing lukis 2

morphemes

2

concepts

1.0

articu

lation index



"What strikes the learner of Malay is the complete lackof those typically Indo-European properties — gender,inflection, conjugation. It is like diving into a bath ofpure logic. Everything is pared to a minimum."

Anthony BurgessLanguage Made Plain

Riau Indonesian

Bislama

Papiamentu

English

Anjing lukis

Dog i dro

E kachó ta pinta

The dog is drawing

2

3

4

6

morphemes

2

2

2

2

concepts

1.0

1.5

2.0

3.0

articu

lation index



0102030405060708090

100

0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3

Articulation Index & BPer Interpretations

0102030405060708090

100

0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3

Articulation Index & BPatPrec Interpretations

0102030405060708090

100

0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3

Articulation Index, Population Size & BPer Interpretations

small (under 100k)

mid-rangelarge (over 100m)

0102030405060708090

100

0.75 1 1.25 1.5 1.75 2 2.25 2.5 2.75 3

Articulation Index, Population Size & BPatPrec Interpretations

small (under 100k)

mid-rangelarge (over 100m)

Conclusions II

• languages with lower articulation indexare more highly associational

• languages with smaller populationsare more highly associational

Question

• why is there so much spreadWithin Malay/Indonesian?

JakartaIndonesian

Riau

Indonesian

SiakM

alay

Kuala Lum

purM

alay

Minangkabau

Kucing

Malay

Bengkulu

PapuaM

alay

Sundanese

Distinct self-appelation

Cannot be referred to asBahasa Melayu/IndonesiaReferred to as bahasa daerah('regional language')Never considered "broken" or"imperfect" languagePrototypically associated withparticular ethnicityMost speakers are of singleparticular ethnicity

Published dictionary

Criteria for Sociolingistic Distinctiveness of Malay/Indonesian Dialects

1st half

JakartaIndonesian

Riau

Indonesian

SiakM

alay

Kuala Lum

purM

alay

Minangkabau

Kucing

Malay

Bengkulu

PapuaM

alay

Sundanese

Distinct conventionalized orthography

Widespread conscious usein newspapersWidespread conscious useon radio

Emblematic use in advertising

Widespread conscious usein popular musicOfficial status as vehicle of instructionin schools

Has its own distinct acrolect

2nd half


JakartaIndonesian

Riau

Indonesian

SiakM

alay

Kuala Lum

purM

alay

Minangkabau

Kucing

Malay

Bengkulu

PapuaM

alay

Sundanese

Sociolinguistic Distinctiveness

Lexical Distinctiveness

Sum: 4S+L

4 3 4 4 0 6 8 14 14

2 24 21 11 14 21 12 23 61

18 36 37 25 14 45 44 79 117


0102030405060708090

100

0 20 40 60 80 100 120

Sociolinguistic Distinctiveness & BPer Interpretations

0102030405060708090

100

0 20 40 60 80 100 120

Sociolinguistic Distinctiveness & BPatPrec Interpretations

Conclusions III

• more sociolinguistically distinct Malay/Indonesiandialects are more highly associational

Why?• two reasons: one "real", one "artifactual"

• the more sociolinguistically distinctive the dialect is,the more appropriate it is to characterize its populationas the relevant subset of the 200m of Malay/Indonesian

• the more sociolinguistically distinctive the dialect is,the smaller its population, and thereforethe higher its associationality

The "real" reason

• this is not a real fact about the different dialectsbut rather an artifact of the experimental method

• the less sociolinguistically distinct the dialect is,the more subjects responses display interference fromthe standard language, which is of low associationality

• in reality, all Malay/Indonesian dialectsmay be of similarly high associationality

The "artifactual" reason

• up to now, all the data was from subjects conforming to

• uneducated

Language-Internal Sociolinguistic Variation

Baseline Sociolinguistic Profile

• low-to-middle class• over 12 y/o• living in community where test language is spoken• tested in same community

… but what about other speakers?

0102030405060708090

100


Baseline Sociolinguistic Profile vs. University Students

-20

-10

0

10

20

30

40

50


Baseline Sociolinguistic Profile vs. University Students

Conclusions IV

• lesser educated speakers are more highly associational

Why?

• the semi-literacy effect:semi-literate subjects latch on to the content words butbut ignore the grammatical information

• the syntactic-mode effect:educated speakers⟺ syntactic modeuneducated speakers⟺ pragmatic mode

• the freak-out effect:uneducated subjects for whom task is unusual panicand simplify their task, resorting to "protolanguage"

• the standard-language interference effect:educated subjects are more likely to display interferencefrom a standardized language of lower associationality

Why Less Educated Subjects Are More Highly Associational

Linguists describing languages based on data fromforeign students (and other expatriates and emigrés)have systematically underestimated the extentto which languages may be highly associational

Badut minum buku

Over 6000 languages waiting forYOU

the association experiment - eva.mpg.de filebare peripheral (bper) zero-marked obliques and other...

Documents